The US Department of War is said to have signed an agreement with Elon Musk's xAI company to integrate the chatbot Grok into secret military systems. This agreement will make Grok the second AI system approved for use on the most sensitive networks of the US military.
Previously, Anthropic's Claude model was the only system operating on secret platforms through cooperation with Palantir Technologies. These networks serve intelligence analysis, weapons development and battlefield operations.
The new move takes place in the context of US Secretary of War Pete Hegseth summoning Mr. Dario Amodei, CEO of Anthropic, to the Pentagon. Accordingly, Mr. Hegseth is expected to request Anthropic to allow Claude to be used for "all legitimate purposes" without applying further restrictive measures, otherwise it may face the risk of being considered a "supply chain risk".
Anthropic opposes removing barriers preventing its technology from being used for mass surveillance of American citizens or deployed in fully automated human-controlled weapon systems. Meanwhile, xAI is said to have agreed with the Pentagon's requests, although the company has not commented.
Sources say Google is also approaching an agreement allowing the use of the Gemini model in a secure environment. OpenAI has not yet reached a similar agreement and is still focusing on developing secure technology.
Pentagon officials admit that replacing Anthropic could cause short-term disruption. Claude was used in the campaign to arrest President Nicolas Maduro last month, marking the first time AI was known to play a direct role in a military raid in progress.
Anthropic builds its image as a company focusing on safety. Mr. Dario Amodei has repeatedly warned about the existing risks if artificial intelligence develops uncontrollably. Last week, Mr. Mrinank Sharma, leader of the company's safety research group, suddenly resigned and issued a warning that "the world is in danger".
The agreement with xAI shows that the US is increasing pressure on technology providers, while expanding AI integration into the most sensitive military systems.
